Show Abstract
Authority identification is an important problem in online information sharing communities such as question answer (Q&A) forums and online social networks (OSNs), where users
care as much about the quality of information being accessed, as its alignment with their interests. This paper investigates a tournament model based approachto authority identification, where interactions between users are modeled as generated by a Bradley-Terry model. We derive a new measure of user authority, the average winnings score, for authority identification in Q&A forums, and evaluate it on data derived from the Stack Exchange Q&A forum. We also show how the log fair bets measure, which has been successfully used for authority identification in OSNs in the past, can be derived from tournament models. We also prove some key results related to a co-ranking framework, for combining information from multiple preference expression graphs based on the same OSN. We then demonstrate the empirical effectiveness of tournament model based approaches, in conjunction with the co-ranking framework.
CITATION
Budalakoti, S., R. Nokhbeh Zaeem, and K.S. Barber, “Tournament Models for Authority Identification in Online Communities,” International Journal of Computer and Information Technology (IJCIT) (ISSN: 2279 – 0764) 6 (2), pp. 75--83, 2017.